By rahulsub
Coding best practices for LLM-assisted development. Based on insights from Karpathy and Anthropic's Claude Code best practices.
Present multiple approaches with explicit tradeoffs before committing to one. Use when there are multiple valid solutions to a problem.
UI feedback loop using screenshots. Provide design mock, implement, screenshot result, iterate until matching. Use for visual work.
Systematic code cleanup after completing features. Use after refactoring or periodically to remove dead code, console statements, TODOs, and tech debt.
Review code like a skeptical senior engineer. Use when reviewing PRs, after generating significant code, or before merging changes.
Manage Claude Code sessions effectively. Use during long sessions, multi-step tasks, or when responses start degrading. Covers /clear, checklists, and subagents.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A collection of coding best practices for LLM-assisted development. These skills help you and your team write better code with Claude Code by applying proven patterns from Karpathy's insights and Anthropic's official best practices.
/plugin marketplace add rahulsub/code-practices-plugin
/plugin marketplace add ~/code-practices-plugin
| Skill | Command | Description |
|---|---|---|
| Explore-Plan-Code | /explore-plan-code | Read files → plan → approve → implement workflow |
| Goal-Driven | /goal-driven | Define success criteria instead of step-by-step instructions |
| Test-First | /test-first | Write failing tests before implementation |
| Naive-Then-Optimize | /naive-then-optimize | Implement obvious solution first, then optimize |
| Skill | Command | Description |
|---|---|---|
| Code-Review | /code-review | Review code like a skeptical senior engineer |
| Simplify | /simplify | Remove over-engineering and unnecessary abstractions |
| Cleanup | /cleanup | Systematically remove dead code and tech debt |
| Surface-Assumptions | /surface-assumptions | Identify and validate hidden assumptions |
| Tradeoffs | /tradeoffs | Present multiple approaches with explicit tradeoffs |
| Skill | Command | Description |
|---|---|---|
| Visual-Iteration | /visual-iteration | UI feedback loop with screenshots |
| Context-Management | /context-management | Manage long sessions with /clear, checklists, subagents |
| Multi-Claude | /multi-claude | Run parallel Claude instances for writer+reviewer patterns |
| Headless-Automation | /headless-automation | CI/CD integration, fan-out migrations, pipelines |
/explore-plan-code
Claude will read relevant files, create a plan, and wait for your approval before coding.
/code-review
Review the changes for hidden assumptions, over-engineering, and edge cases.
/naive-then-optimize
Implement the obvious correct solution first, then optimize while preserving behavior.
These practices are based on:
MIT
npx claudepluginhub joshuarweaver/cascade-code-general-misc-2 --plugin rahulsub-code-practices-pluginBehavioral guidelines to reduce common LLM coding mistakes, derived from Andrej Karpathy's observations on LLM coding pitfalls
Six Karpathy-inspired principles that stop Claude Code from making silent assumptions, overengineering, doing drive-by refactoring, and patching bugs blindly.
English-first 12-rule behavior contract for Claude Code, Cursor, and agent skills
Intelligent prompt optimization: injects the right context at the right moment so Claude lands a better first output. Clarifies vague prompts with research-based questions, plus targeted nudges for approach selection, plan readability, workflow routing, background execution, subagent routing, output readability, user-decision questions, and plan-mode assessment
Claude Code workflow patterns: prompting, CLAUDE.md maintenance, multi-agent orchestration
Sets up your project for effective AI-assisted development — optimized CLAUDE.md files, effective coding guidelines, formatter hooks, built-in code-quality agents, TDD and test coverage, all tailored to your actual codebase.